There is a problem that causes sites not to render if they included orphaned pages. @robatron fixed this, but I reverted the change in the pagination sprint. Figure out how to make the two things play together.

It was convenient for reading in the header data, but now it complicates the external interface. Everything should be a property of Page.

How can it still support arbitrary metadata being passed if this happens?

Refactor the data structures used to store sub pages to make it suck less.

Today's... fun... with regression, and errors and github/git being wierd for me has inspired me that tests are needed.

Possible variables:

• date stuff (year, month, day, day of week)
• author
• arbitrary variables? (kind of crazy, but it might be cool)

When you publish a post the author doesn't show up correctly in the output html. I have a local fix to this but I doubt it's what really should be done to fix it.

uberj@f00a37e

Many time it would be convenient to have a "preview" or partial version of a page that is user controlled.

I propose that these two versions both work, and be equivalent:

Option 1 - Excplicit

title: Foo
preview: This is the preview of Foo.

---
This is the preview of Foo.

This is not the preview of Foo.

Option 2 - Implicit

title: Foo

---
This is the preview of Foo.

---

This is not the preview of Foo.

In both cases the content of Foo would be

This is the preview of Foo.
This is not the preview of Foo.

Currently wok assumes that it is generating html files. It should stop doing that.

It would be super-fly to have a simple content tagging system that would allow us to:

• Tag content with a comma-separated list (e.g., tag: foo, bar, herp, derp)
• Access tag list in the templates with tags
• Access tagged pages in the templates with tags['foo'].pages[0]

Both in the template interface, and the yaml header, date should be the date only (no time), time should be the time only (no date) and datetime should be the datetime.

Be smart about when multiple are specified, combining them. Page.date and Page.time should probably call out methods of Page.datetime to extract the relavent bits.

All attributes of page exposed to the templates should have good str() methods. This way when they are printed in template they look good.

For some reason python 3 doesn't like unicode string notation.

Traceback (most recent call last):
  File "setup.py", line 5, in <module>
    from wok import version
  File "/home/uberj/wok/wok/__init__.py", line 1
    version = u'0.8.0'
                     ^
SyntaxError: invalid syntax

uberj@francium ~/wok (> ^ ^)> python --version
Python 3.2.2

Things like post specific images. Will live in the content directory, along side the content, but will be treated similar to media in the media directory.

Further to our IRC chat yesterday - I can confirm the pagination is still broken whenever only 2 pages are generated.

My previous example was 2 files and a limit of 1 - I'm now using 9 files and a limit of 5, and next/prev links don't work. The prev link works if I manually go to page 2, as you demonstrated yesterday.

This is using v0.8.1 stable from PIP.

Markdown and restructured text should be used when available, but not be hard requirements.

@robatron had the idea to have arbitrary code that could be ran to generate data for the site. This might be a good way to make wok extensible.

In particular, the template variables added by the Page class aren't in the dictionary, because the hooks is called from the engine, not from the page.

@MostAwesomeDude and @robatron have had this issue, but I can not reproduce it.

The sample site should be the documentation site. It should show off the features of wok in an attractive way, both the generated pages, and the source code itself.

The current sample site is ugly, and doesn't make much sense to anyone except me.

List index out of bounds exception

This means we won't clear out the output directory, and we have to be a lot smarter about things.

The vary-by feature would regenerate a page many times for every value of a particular variable. For example a page with the metadata vary-on: site.tags would be generated once for every tag on the site, and would have a variable that contained the name of each tag.

This would be good to make things like archive pages or to browse by tag.

In particular, times are usually thrown away. Fix it!

Have a {{ site.categories }} template tag that returns a list of categories and their associated pages, e.g., {{ site.categories['blog'] }} would return all pages with the category 'blog'.

The author metadata field should allow multiple authors to be specified, in the case of collaborative editing.

Mainly for index pages.

How would this work? Right now I am thinking of this: Somehow a page will indicate it wants to be paginated, how many items there are per page, and someone give a collection of the items that make up a page. Then, during rendering, the current page number and other related info will be available as a template variable.

As the confusion above hints at, wok is not well suited to this.

Wok trys to read Vim swap files as content/templates. Results in an error similar to this:

Warning from metadata: You didn't specify a title in  .index.mkd.swp. Using the file name as a title.
Traceback (most recent call last):
  File "../wok/wok/wok.py", line 82, in <module>
    Wok()
  File "../wok/wok/wok.py", line 26, in __init__
    self.render_site()
  File "../wok/wok/wok.py", line 78, in render_site
    p.render()
  File "/home/rob/wok/wok/page.py", line 157, in render
    self.html = template.render(templ_vars)
  File "/usr/lib/pymodules/python2.7/jinja2/environment.py", line 891, in render
    return self.environment.handle_exception(exc_info, True)
  File "templates/default.html", line 7, in top-level template code
    {{ page.content }}
UnicodeDecodeError: 'ascii' codec can't decode byte 0xa3 in position 16: ordinal not in range(128)

The render content setting should control whether or not a content file gets written to disk. It should not affect whether or not it gets rendered, templated, and passed to other bits of the code, like index pages and fragment-generated pages.

Tsk, tsk ;-)

C:\Users\ew630d\Documents\mystuff\wok>python -m wok.engine
Traceback (most recent call last):
  File "C:\Python27\lib\runpy.py", line 162, in _run_module_as_main
    "__main__", fname, loader, pkg_name)
  File "C:\Python27\lib\runpy.py", line 72, in _run_code
    exec code in run_globals
  File "C:\Python27\lib\site-packages\wok\engine.py", line 171, in <module>
    Engine()
  File "C:\Python27\lib\site-packages\wok\engine.py", line 51, in __init__
    self.load_pages()
  File "C:\Python27\lib\site-packages\wok\engine.py", line 115, in load_pages
    self.options, renderer))
  File "C:\Python27\lib\site-packages\wok\page.py", line 53, in __init__
    self.build_meta()
  File "C:\Python27\lib\site-packages\wok\page.py", line 90, in build_meta
    util.out.warn('Your slug should probably be all lower case, and '
  File "C:\Python27\lib\site-packages\wok\util.py", line 19, in warn
    self.pretty_print("Warning: {0}".format(message))
  File "C:\Python27\lib\site-packages\wok\util.py", line 31, in pretty_print
    _,w = [int(n) for n in os.popen('stty size', 'r').read().split()]
ValueError: need more than 0 values to unpack

Normally I'm not one to complain about Windows compatibility issues, but I'm trying to show that wok will work well in a Windows environment to grow the user base :)

This makes parsing a bit nicer, and will prevent confusing people who already understand YAML.

Old style CSV tags (or other such listed values) should still work, but give a deprecation in 0.7.x.

Use setup.py

It would be nice to generate RSS feeds from a wok site. To do this the template engine needs to stop assuming it is making html files all the time, and pages will need to provide a little more metadata. In particular, they need a unique id that won't change between runs.

When I pip install wok, it installs all of the dependencies except for pygments, which I have to install via aptitude.

wok --server should generate the site then launch a webserver server on a port that is not 80 to test the site quickly and easily.

Integrating something like latex math would be really cool for making pretty equations. This should probably be made with some sort of extensible plug interface.

It isn't obvious that wok succeeds when it runs. There should be some sort of output to signal that all went well.

What kind of things should that output say? List of files generated? Number of files?

Unicode in content files should be automatically escaped to html elements, such as acute "e"s.

@MostAwesomeDude and I were talking about Wok's category system, and I have a suggestion: I don't think categories should have "parents", i.e., special content that has access to page.subpages. Instead, I suggest that every member of a category should have access to every other member of that category, i.e., pages.siblings.

This would solve a couple big problems:

• Wouldn't force the user to set the slug as 'index', which is not intuitive
• Wouldn't make the user keep track of which content is the parent of what category

It appears that the site.content.gather.post hook is not providing its attached functions with the pages variable like it's supposed to according to the hook docs.

For example, in my function attached to the hook, I'm including a pages parameter, but when I try to run wok, I get the following error message:

rob@vinux ~/robmd.net $ ./wok --server
Traceback (most recent call last):
  File "/home/rob/wok/scripts/wok", line 4, in <module>
    Engine()
  File "/home/rob/wok/wok/engine.py", line 103, in __init__
    self.load_pages()
  File "/home/rob/wok/wok/engine.py", line 242, in load_pages
    for pages in self.run_hook('site.content.gather.post'):
  File "/home/rob/wok/wok/engine.py", line 167, in run_hook
    returns.append(hook(*args))
TypeError: process_title_images() takes exactly 1 argument (0 given)

It might be nice to be able to set a global author in the site config file that can be overwritten by the content-specific author field. That way, one would't always have to enter an author for every blog post, for example.

Regardless of if I include the 'date', 'datetime', or 'time' field or not, I'm getting the following key error:

Traceback (most recent call last):
  File "../wok/wok/wok.py", line 82, in <module>
    Wok()
  File "../wok/wok/wok.py", line 24, in __init__
    self.load_pages()
  File "../wok/wok/wok.py", line 61, in load_pages
    self.all_pages.append(Page(os.path.join(root,f), self.options, renderer))
  File "/home/rob/wok/wok/page.py", line 80, in __init__
    self.build_meta()
  File "/home/rob/wok/wok/page.py", line 143, in build_meta
    self.meta['datetime'] = isodate.parse_datetime(self.meta[datetime_name])
KeyError: 'date'

Wok fails if the 'media' directory doesn't exist. If the media directory's only purpose is to contain media files, shouldn't it be optional?

Traceback (most recent call last):
  File "../wok/wok/wok.py", line 82, in <module>
    Wok()
  File "../wok/wok/wok.py", line 23, in __init__
    self.prepare_output()
  File "../wok/wok/wok.py", line 43, in prepare_output
    for name in os.listdir(self.options['media_dir']):
OSError: [Errno 2] No such file or directory: 'media'

The publish metadata setting should make a file never show up in the variables of any template. As it is now a page with publish: false would still appear in a blog index page, but would not appear on disk, so would create a broken link.

Use pygment to do code highlighting. Preferrably there would be a way native to each type of renderer to mark a chunk of text as code. Ideally you could also mark this code as a particular language.

Having an option to re-render the site when files change will make developing sites quicker and more convenient.

Tell me if I'm being stupid, but site.date and site.time don't seem to be set - I've looked through the code briefly and assumed it should be set here https://github.com/mythmon/wok/blob/master/wok/engine.py#L298

Would also be nice to not have microseconds appended to site.datetime unless there's a good reason for it? Maybe consider adding the timezone instead/as well. Let me know if you'd like me to raise those as separate feature requests :-) Edit: Just noticed you're using site.datetime.strftime() for this!

Sample case was this:

=====================================
Something that should have been an h1
=====================================

Some text

The header (wrapped in ===) did not show up.

Add options for the structure of the generated sites.

• Classic - /category/subcategory/slug.html (default, implemented now)
• Clean - /category/subcategory/slug/index.html (possible now, but could be done cleaner)
• Flat - /slug.html (For super simple sites)
• Manual - Force all pages to define a url category.

I feel like I had another idea for an option, but I don't remember what it was now.

Pages should have a url field. This can be defined in the YAML header to place the file in a non-default location. From templates it can be accessed to make links to pages.

It should be a relative path from the root of the output dir.

Custom variables (variables beside the guaranteed ones) get propagated from a page to any page processed after it.